Welcome to the ATTD 2023 Interactive Program
IS010 - Exercise with Automatic Insulin Delivery (ID 189)
Abstract
Abstract Body
Physical exercise with type 1 diabetes is a challenge, regardless of whether the insulin is given as multiple daily insulin injections or insulin pumps. Current consensus guidelines for avoiding hypo- and hyperglycemia during and after exercise are available. Recent advances in diabetes technology have led to the development of automated insulin delivery (AID) systems for glycemic management of people with type 1 diabetes. However, little is known about their safety and efficacy around exercise, which can cause significant and often worrisome disruptions in acute glycemic control. Although consensus recommendations exist for exercise management with AID, the guidance is based on first-generation AID systems. Therefore, it is unknown how best to use the latest diabetes technologies around exercise.In this presentation, data from new and ongoing studies that have investigated different glucose management strategies for training with the different generations of AID systems will be discussed. In addition, possible future management options for spontaneous exercise during AID treatment will be discussed.
IS011 - Is technology useful for breaking down barriers to exercise in diabetes? (ID 190)
Abstract
Abstract Body
Clinical exercise guidelines recommend that children should aim to achieve at least 60 minutes of moderate-to-vigorous physical activity (MVPA) daily, but many youths with type 1 diabetes (T1D) are falling short of these recommendations. For individuals with T1D, exercise and physical activity can lead to disturbances in glycemia without proper preparation and implementation of these strategies. Some common strategies include insulin dose adjustments and/or carbohydrate feeding to reduce the risk of hypoglycemia during exercise.
In addition to barriers such as lack of motivation to exercise and fear of hypoglycemia, exercise interventions in adults with T1D have been shown to be acceptable and feasible to deliver. Our team at Stanford is currently implementing a structured telehealth exercise education program in newly diagnosed youth with T1D. The exercise pilot study is part of the larger Teamwork, Targets, Technology, and Tight Control 4T Study that started youth with new-onset T1D on continuous glucose monitoring (CGM) technology, physical activity trackers, and exercise education approximately 1-month after diagnosis.
This study also examined the potential association between physical activity and glycemia on active days in youth with T1D. We present data from focus groups aimed at understanding the parental and youth experiences in exercise education after T1D diagnosis and also benefits and challenges with real-world use of physical activity trackers.
IS012 - Physical activity with long and ultra-long-acting basal insulins (ID 191)
Abstract
Abstract Body
The use of long and ultra-long acting basal insulins exposes people who are physically active to insulin levels that are entirely different from normal physiology. Consequences involve exercise-induced hypo- and hyperglycaemic excursions, alterations in substrate utilization and post-exercise metabolism. Large variations in individual clinical needs (e.g. purpose for engagement in exercise) introduces additional complexity. However, several pro-active strategies, including variation of exercise intensity, use of digital tools, pharmacological agents and nutritional strategies can help people on long and ultra-long acting basal insulins achieving maximal benefits from being physically active.
IS013 - Automated detection of meals and exercise events in people with diabetes (ID 192)
Abstract
Abstract Body
Background and aims
The occurrence of several factors that disturb glucose homeostasis must be detected In real time to develop fully-automated insulin delivery (AID) systems. Meals, physical activities (PA) and acute psychological stress (APS) events should be detected and their characteristics should be determined to compute the optimal insulin doses to be infused by the AID system or suggested to a user in an advisory system.
Methods
Machine learning techniques ranging from support vector machines and decision trees to qualitative trend analysis and deep neural networks, along with systems engineering and multivariate statistical techniques can detect the occurrence of specific events, discriminate between different types of events and estimate the characteristics of the specific event (carbohydrates in a meal, intensity of PA, type of APS). Processing of signals from CGMs or wearable devices to filter out signal noise and motion artifacts, imputation of missing data, generation of features from measured variables and pruning of features to minimize collinearity in information improve detection and diagnosis accuracy.
Results
Various techniques can infer events that have occurred from CGM values reported (yielding feedback information) or from wearable device data as the event is occurring, well before it affects CGM readings (feedforward disturbance information that will affect glucose levels). Estimates of carbohydrates in a meal based on CGM data can provide miniboluses of insulin during a meal, Detection of physical activity type and estimates of energy expenditure inform the control algorithm of the AID system to enable adjustments of insulin infusion doses. Discrimination between PA and APS prevent incorrect dosing of insulin based on erroneous assumption that an increase in heart rate would always indicate PA.
Conclusions
The quest for fully-automated AID systems benefit from leveraging real-time data provided by wearable devices such as activity wristbands. Signal processing, systems engineering and machine learning techniques that can work with data generated in free living must be used for generating reliable information for use by the AID system.
Q&A (ID 193)